Introduction

The 2013 global tuberculosis report of the World Health Organization (WHO), based on vital registration data, estimated that TB caused 74,000 deaths in HIV-uninfected children globally in 2012 [1]. Recent data suggest that TB in under-15 children may contribute 10–20% of the total disease burden in endemic countries [2,3]. Due to lack of reporting of TB as a cause of deaths in HIV and/or pneumonia related deaths [1,4,5], the actual burden of childhood TB is likely to be higher than these estimates. However, the main impediment to understand the actual burden is the difficulties in the confirmation of diagnosis of TB in children. Obtaining sputum samples from young children can be challenging and disease is paucibacillary, so yield from microscopy is low. Mycobacterial culture takes as long as 8–12 weeks for a result in conventionally used solid cultures, and while real-time PCR technique such as Xpert MTB/RIF assay is much quicker, sensitivity is suboptimal compared to culture [6], so a negative result does not rule out a diagnosis of TB. Undoubtedly, there is a huge need for an accurate and rapid diagnostic test for TB in children.

Our group has previously published novel data of encouraging results from evaluating a blood-based test measuring antibodies in lymphocyte supernatant (ALS) for diagnosis of TB in adults and children [7,8]. The study in children was limited to those with a clinical diagnosis of TB. Among 58 children with clinically diagnosed TB, 9 (15%) had culture-confirmed and 53 (91%) had ALS positive TB [8]. Further, the levels of ALS may be affected by reduced immune function such as occurs in children with severe malnutrition. We recently conducted a prospective study of the prevalence of TB among children with severe malnutrition and radiologic pneumonia [9]. In a subset of children from that study, we have evaluated the diagnostic performance of ALS in the diagnosis of TB in comparison with culture and Xpert MTB/RIF in addition to comparison with clinical diagnoses of childhood TB.

Materials and Methods

Results

Among the 405 children enrolled during the study period [8], ALS from a blood sample was measured in the first 224 children. Of these, 221 children had gastric lavage fluid and induced sputum obtained for microscopy and mycobacterial culture, and 51 of these children also had gastric lavage fluid and induced sputum analyzed by Xpert MTB/RIF. Only one sample of gastric lavage fluid and induced sputum was taken from each study patient and microscopy, mycobacterial culture, and Xpert MTB/RIF were done both from the gastric lavage fluid and induced sputum of each patient. The median (IQR) age of the study children was 10 (5, 16) months and age range was 1 to 59 months. Of the 221 study children, there were 12 (5.4%) children with microbiologically confirmed TB—8 by culture and 5 by X-pert MTB/RIF (1 positive in both tests)—and there were 41 (19%) with clinical, non-confirmed TB and 168 that were considered not to have TB.

Overall, ALS was positive in 89/221(40%), negative in 85/221 (39%), and borderline in 47/221 (21%) children, and the proportion that were positive, negative or borderline are listed in Table 1. The proportion of positive ALS was found to have no significant difference among study children who had “Confirmed TB”, “Non-confirmed TB”, and “Not TB” (Chi-Square for linear trend for positive ALS from “Confirmed TB” to “Non-confirmed TB” to “Not TB” = 1.098, p = 0.295). The clinical characteristics of the severely malnourished children with pneumonia are compared between those with and without a positive ALS in Table 2. The characteristics of the group of children with a positive ALS result were similar to those of the group of children with a negative ALS result, although the former were significantly older on average, and a significantly greater proportion was BCG-vaccinated. The sensitivity, specificity, positive predictive value, negative predictive value and accuracy of ALS at different cut-offs for “confirmed TB” versus “not TB”; and for “all TB” versus “not TB” are shown in Table 3. Although, the sensitivity of ALS with the lower cut-offs was increased from 67% to 75% for “confirmed TB” versus “not TB” and from 57% to 66% for “all TB” versus “not TB”, however, the specificity and accuracy of ALS with the lower cut-offs gradually decreased from 51% to 39% and 52% to 41% for “confirmed TB” versus “not TB” and 51% to 39% and 52% to 45% for “all TB” versus “not TB—respectively. Baseline characteristics of severely malnourished children according to diagnostic groups (confirmed TB”, “Non-confirmed TB”, and “Not TB”) are shown in Table 4.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary results of ALS in blood (using an optical density ≥ 0.60 as a positive cut-off and < 0.38 as a negative cut-off), confirmed TB, non-confirmed TB, and not TB in children under five with suspected pneumonia and severe malnutrition.

https://doi.org/10.1371/journal.pone.0126863.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Characteristics of severely malnourished children with suspected pneumonia who had their ALS done.

https://doi.org/10.1371/journal.pone.0126863.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Validity of ALS at different cut-offs for “Confirmed” versus “Not TB”; and for “all TB” (“Confirmed” plus “Non-Confirmed”) versus “Not TB”.

https://doi.org/10.1371/journal.pone.0126863.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Baseline characteristics of study participants according to diagnostic groups.

https://doi.org/10.1371/journal.pone.0126863.t004

Discussion

The important observation of this study is the poor performance of ALS in the diagnosis of TB in our study population. This test was originally validated in an adult population and found to have a sensitivity and specificity of 92.5% and 80% compared to smear positive and microbiologically confirmed TB, respectively [7]. Also, in a previous study investigating ALS in children with clinically diagnosed TB (age range: 11 to 167 months) the sensitivity and specificity of the test were 91% and 87%, respectively [8]. However, in this population of severely malnourished and young children, the diagnostic accuracy of the ALS was poor including in children with microbiologically confirmed TB.

The ALS test is based on the assumption that antigen-specific plasma cells are present in the circulation only at times of active TB disease and not in asymptomatic infections i.e. in latent TB infection [8]. ALS is thus expected to have greater specificity for active disease because it quantifies the amount of antibody actively secreted by TB-specific plasma cells [11]. Based on this, ALS could be expected to perform well in cases with microbiologically confirmed TB. However, among the study children who were considered not to have TB (i.e. who did not have either confirmed or non-confirmed TB), 38% of those children were observed to have a positive ALS result, which either indicates that there are significant issues with test specificity or alternatively could indicate that the ALS assay also produces positive results in children with latent TB infection who coincidentally have another respiratory infection. Notably, none of the children in the “not TB” group had a positive TST result, but it is well-documented that false-negative TST results are common in children with severe malnutrition [12,13]. Due to financial constraints we were not able to perform interferon-gamma release assays (IGRAs), which could have further contributed to excluding potential latent TB infection in those children. One possible explanation for the poor performance of ALS in our study population is that both cell mediated and humoral (i.e. plasma cell function) immune responses are substantially impaired in children with severe malnutrition [14,15]. It is worth noting that our study participants were on average considerably younger than the participants of the two previous studies that evaluated ALS. Similar to other tests that rely on immune response, it is possible the immaturity of the immune system in our study participants has resulted in an increase in the proportions of negative or borderline results, a phenomenon also observed with IGRAs [16,17,18].

We have considered the performance of ALS at different cut-offs. Although, at lower cut-offs the sensitivity of ALS improved slightly, simultaneously false positivity of ALS was increased, thus, resulting in reduction of the specificity and accuracy of ALS that potentially make the performance of the test worse.

The main limitation of the study is a small number of children with microbiologically-confirmed TB that may hinder the actual ability to assess test accuracy. The vast majority of cases were classified as ‘non-confirmed’ TB, some of whom might not have had TB. This may lead to data contamination, which might make test performance appear worse than it actually is. However, our approach to assess the performance of the ALS by comparing children with ‘only confirmed TB’ versus ‘not TB’ intended to address this limitation.

In conclusion, measurement of ALS does not appear to have a role in the diagnosis of TB in severely malnourished children in our population.

Acknowledgments

We gratefully acknowledge the donors for their support and commitment to icddr,b's research efforts. We would like to express our sincere gratitude to Professor Stephen M Graham for his expert review in improving the quality of the manuscript. We would also express our sincere thanks to all physicians, clinical fellows, nurses, members of feeding team and cleaners of the hospital for their invaluable support and contribution during patient enrollment and data collection.